← All episodes
Alex KeeslingNate Gemelke

Episode · Nov 10, 2022

Alex Keesling & Nate Gemelke, QuEra

Why analog neutral atoms may reach fault tolerance first

Summary

Alex Keesling, CEO, and Nate Gemekle, CTO of QuEra, a company building quantum computers based on neutral atoms with a unique analog quantum computation mode, are interviewed by Yuval Boger. Alex, Nate, and Yuval discuss why QuEra chose a different path toward universal, fault-tolerant quantum computers, the significance of recent qubit shuttling experiments, what they learned from the Amazon Braket integration, and much more.

Key takeaways

  • QuEra bets on analog Hamiltonian evolution alongside gate-based digital mode on the same neutral atom hardware, arguing this dual approach gives useful results before full fault tolerance arrives
  • Recent qubit shuttling experiments, where atoms physically move between zones for entangling gates and readout, are a key building block QuEra needs for error-corrected, fault-tolerant machines
  • Running on Amazon Braket exposed QuEra to customers outside physics labs, and most of them cared about solving a business problem rather than which qubit technology powered it
  • Neutral atom hardware's reconfigurable qubit layout is presented as a structural advantage for implementing error correction codes compared to fixed-connectivity architectures

Read the full transcript on the Quantum Computing Report site here

Transcript

Yuval Boger: Hello Alex, hello Nate. Thanks for joining me today.

Alex Keesling: Hey Yuval. Thank you for having us. It's a pleasure.

Nate Gemelke: Hey Yuval.

Yuval: So Alex, who are you and what do you do?

Alex: I am Alex Keesling. I'm the CEO of QuEra Computing. We're a Boston-based startup building quantum computers with neutral atoms and the rest of the stack that goes with it.

Yuval: And Nate, how about you? Who are you and what do you do?

Nate: You're starting with such deep and difficult questions. I'm QuEra's CTO.

Yuval: All right. My understanding is that QuEra is building a quantum computer based on neutral atom technology. There are obviously other companies who are using neutral atoms for their quantum computers.

What's different about your approach?

Alex: Yeah, I think that's a great question. As you say, we are building quantum computers using neutral atoms. Our atom of choice for this is rubidium. It's something that we have carried over from QuEra's beginnings as an effort that spun out of many years of research at Harvard and MIT, where we developed this platform, and we found rubidium to be an excellent candidate for this.

So this is one of them. The other thing is that we're really focused on creating efficient tools for quantum computers that provide the most useful value, both now and at every point along the evolution of the technology. So our first generation of machines is using an analog computational mode that is different from many of the other quantum computing players.

We are using fewer controls, but going to larger system sizes and particularly focusing on finding the applications that have an efficient encoding in the hardware as it exists. So we leverage the flexibility of being able to effectively reprogram the connectivity of our quantum computer every time we run a program.

Yuval: So Nate, maybe you could help decompose this for me.

So first, congratulations. I think you just launched the computer on AWS Braket, and it has, I think, 256 qubits. Are those qubits as in gate-based qubits? What does analog mode mean, and what kind of problems can it solve?

Nate: Yeah. Obviously this is a new machine, and you do have to understand its strengths and weaknesses pretty carefully.

We're calling these things analog Hamiltonian simulators right now. And the way that this machine works is you can put any number of atoms that you want, up to that 256, and position them anywhere that you want on a plane, and set up a calculation with a new register basically every time you run a calculation.

And beyond that, obviously you want to process quantum information. So the way that we do that is we drive the atoms between states: the ground state of the atom and the excited Rydberg states. And when they're in the Rydberg states, they like to interact when they're close by. So that's essentially the mode of operation of the machine: you program the position of the atoms, you fix them while you run a pulse sequence, and you run a calculation that way.

So it's a little bit different from gate-based computing, where you're really running a unitary targeted on single qubits and pairs of qubits at a time. In a way, you're doing operations on the whole array at once, and using this programming of the positions and the timing of pulse sequences to your advantage.

It's a really great tool for doing a few things that we know about today, and probably some things that we'll find out about as people log on and start playing with it and trying things that even we didn't expect. But we do know that it's really good for doing things like quantum simulation, like looking at materials problems and other kinds of quantum systems, and trying to simulate what they do by arranging the atoms in the right way and doing pulse sequences in the right way.

It's also really good for optimization problems, like finding independent sets in graphs. We have some work on that, and we have exposed the computer on the cloud in a way that you can run algorithms for both of those kinds of things.

Yuval: Let's assume for a second that you are not with QuEra, but you are advising a commercial customer. So a commercial customer has an AWS account, goes to Braket, and sees a whole bunch of quantum computers: D-Wave, IonQ, and QuEra. When should they use QuEra? That's question number one. And two, how do I program the machine?

Is it just Qiskit, like I program an IonQ machine, or is it something else?

Nate: I guess that one's to me. Yeah, so being that it is not a gate-based computer, Qiskit is not a particularly great fit for it. We have a relatively simple intermediate representation that you can use to program the machine, and it's really based around what the machine can do. So what you'll find there is an ability to move the atoms around in space, and pulse sequences in time, and that's kind of the natural mode of operation for the machine.

Now, over time, we will improve the capabilities of these machines, and we will roll out new features that look more and more like gate-based machines as well. And as that happens, I think you'll see a melding of the languages that you use to program it.

Yuval: Do you see the future as hybrid analog-digital quantum, or is it ultimately just going to be a fault-tolerant universal gate-based machine?

Nate: Well, I think everybody is waiting for the day that it's a fault-tolerant machine, and we're not just waiting, we're trying to make it happen here.

So ultimately, I do believe that's what we'll see. But in a lot of ways, a hybrid of analog and digital is probably a step along the way. There are a lot of different meanings behind the word "hybrid," depending on what different things you bring together.

There's also hybrid computing, where you bring classical resources and quantum resources together. I think that will be a step along the way. And as those things become successful, I doubt that we will leave them behind after we've seen success with them. So in the end, I doubt that quantum computing will be exactly what everybody imagines today.

I think there will be surprises along the way. I think that's the best reason to build computers now and get them online, because we're going to find those surprises. Having the vehicles to do that and putting them in basically the whole world's hands, that's the recipe for success for us all.

So let the surprises begin. Once they begin, I don't think that we're going to leave them behind either. So I think we're going to see aspects of all of those things at the end of the day, when we really have these machines in our daily lives.

Yuval: Alex, I think you were part of the original team at Harvard that developed a lot of this technology, and I saw some recent articles about qubit shuttling. Could you put it in context of where this kind of technology would lead us?

Alex: Yeah, absolutely. I must say that it was a very exciting time, when I was doing my PhD, to be developing this platform and to really see the amount of progress that we were able to make very quickly. We started talking just now about analog processing and Hamiltonian simulation, and this is one of the strengths of the platform: the fact that we can move atoms around, even just to initialize the connectivity of the processor.

And this has allowed us to encode problems very efficiently. Now, the next step in the technology is taking that idea further. The very exciting thing about the recent results that you're describing is that they have allowed us to look at this platform as a viable way to also do gate-based computations, in a way where part of the programming is to reconfigure the chip in real time, meaning that we can encode information in the internal state of our atoms and then move them around to reconnect them with other ones.

In the published work, this was a way to very efficiently implement things like quantum error correction codes. Just using these tools, we saw that it is possible to move very quickly between different kinds of applications, including a hybrid analog-digital approach, where you do some part of the computation in an analog way and then do post-processing using gates.

You can even do pairwise interactions between atoms that were far away during the analog evolution. And taking these ideas further, I think, shows us that there's a path to scalability. When we talk about neutral atoms, we very often talk about how easy it is to increase the number of atoms in the processor, given that they don't necessarily repel one another, and everything is done using lasers.

But there's another consideration to scaling up the size of quantum computers, which is how many controls you need, and effectively how many controls you need per qubit. What we're looking at in terms of the evolution of the platform is to continue to increase the number of qubits that we can use within this gate-based approach, but, by virtue of being able to move atoms in and out of different regions, to keep the number of necessary controls low as we continue going to larger and larger useful qubit counts.

Yuval: And does that also allow you to do essentially mid-circuit measurements: move qubits to a different region and just measure a few of the qubits without interfering with the rest?

Alex: Absolutely. That's one of the things that we are looking to enable: having a region that allows us to read out the ancillary qubits, for example, for quantum error correction protocols.

And also the ability to bring fresh qubits into the computation.

Yuval: Nate, I wanted to ask you, why aren't other neutral atom companies following the same path? Does the path of delivering analog quantum computation perhaps make it longer for you to get to the holy grail, the universal gate-based fault-tolerant computer, or do you see it differently?

Nate: I don't want to speak for others, but I will say that putting out actual products that are analog now allows us to take a pathway to the ultimate goal, which, as we talked about before, is fault-tolerant quantum computation, on a different pathway than I think most people think about on a daily basis.

Usually people work to make small systems that are fully programmable, and then scaling is a game of making a fully programmable computer larger and larger. What we've chosen to do is a little bit different. If you think about our pathway on a plane, one direction signifies the programmability of the device and the other one signifies the scale.

We've gone to make the scale as large as possible first, and then improve the programmability later. What that lets us do is put out a first product where, very clearly, there's nothing on earth that can tell you what it's going to do. You just have to run the machine itself, and that's something special.

That means that the machine is unique on earth. And at every step along our pathway, we can continue to make a machine where you just can't predict what it's going to do. And that's what the age, or the era, of discovery really means in a technology arc: you're really learning new things with every product that you release.

And in this case, we're learning that on behalf of the world. So I think it's an exciting pathway. It's definitely the right pathway for me personally, because I like to take my cues from nature about how to build the right technology, and the only way to do that is to ask nature itself.

Yuval: So that sounds a little bit like the Feynman quote, right?

Or was he referring to gate-based machines?

Nate: Well, honestly, my personal read of what Feynman said… and I'm definitely an older guy than Alex is, but I wasn't there to hear it in person. I've only read it. But my read on what Feynman said is that it's very close to the heart of building quantum simulators.

And in fact, there's a line in what he said at one of his lectures that basically describes a neutral atom machine verbatim: representing every qubit with a single spin in an atom. So I'm proud to have delivered on the dream that I wasn't there personally to hear, but could read in print.

Yuval: Alex, could you tell me a little bit about the company? How large is the company? How much funding? When was it started? And anything else you can share.

Alex: Yeah. As I was alluding to earlier, the company is the result of work at Harvard, where in 2019 we had accumulated enough evidence that these neutral atom systems were powerful enough to deliver on the promise of having quantum systems that could do something useful, at least for us at that time.

And that was when the company was put together. Since then, the company has been structured in a way that really takes a holistic view of quantum computing and allows us to work out the ways that we can extract the most value out of near-term quantum computers as we build towards the longer-term goal of fault-tolerant quantum computers.

And that's the reason why the company has people working on hardware, initially commercializing the developments from the universities, but now developing them further. We have an engineering team that is taking these machines to a different level than what has been done in academic settings.

We have, of course, a drive to bring these machines to customers, to provide something to them, and to learn from them. And we have a group of software developers that are building tools for customers to both access the machine and learn how to program it and get the most out of it.

And last but not least, we have a group of application scientists who are interfacing with customers and working with them on how we take the problems that they have and convert them into things that we can run on the hardware now, and how we can reduce the overhead in implementing all of these algorithms.

This, we think, is crucial, especially for the next few years. Going back to what Nate was saying, we really want to discover the best applications for quantum computing. And for that, we need to be very conscious of using the quantum resources that we have efficiently.

So that forms the full stack. At this point, we have about forty people. We've been very successful at growing a fantastic team of very high-quality people who are all motivated to bring this dream of useful, practical quantum computing to fruition.

Yuval: When you think about the AWS Braket integration, now that you're on the other side of it, what advice would you give to someone who's just built a computer and wants to go through that process?

Alex: Oh, there's a lot to say there. But something that we always felt was very important was to keep in mind who the customers are going to be, who's going to be using the machine and how. And really focus on the use cases that we expect to see, at least in the early stages of access to the technology, by talking to potential customers, and develop things like the interfaces and the intermediate representation that Nate discussed with that in mind, by getting constant input from others.

That, of course, comes with working closely with the team on the other side to make sure that we are working together to define standards, to always put the customer first, and to figure out how we align on connecting our part of the technology stack with the service that they're offering. And also working together on how we're going to get our message out there, and how we're going to make sure that we're letting people make the best use of our machines.

Nate: Great. Alex left out figuring out where to get pizza at 3:00 AM.

Yuval: How do you feel about the state of quantum hype? Do you think there's too much hype? Do you think there's not enough of it? Where does QuEra stand on the hype scale?

Alex: I'm very excited to see how much excitement there is behind quantum computing, because I think that it will be an absolutely revolutionary technology.

Recently, I've seen more of an effort to focus on benchmarking and on really creating clear expectations of where the technology is and where it's going in the future. I think that there's a danger in hype getting out of control, and here at QuEra, this is something that we think about constantly, right?

We want to make sure that people understand the promise of quantum technology, but that they also understand where we are right now in the stages of technological development. There are absolutely applications where we expect that these machines, the existing machines, are going to provide something very valuable, in understanding how to program them and understanding how the evolution of these machines is going to bring more and more power to computational tasks.

But I think that we still need to work very hard on doing the right matching of existing capabilities to the problems that we want to solve. And for us here at QuEra, this is exactly what led to the decision to have our first machine work in an analog processing mode. Because we have seen, from the work that university groups around the country and the globe have done, that this is a very powerful way of using quantum resources, in a way that has the power to provide answers and insights that are very hard to replicate with classical solutions.

Yuval: So as we get close to the end of our conversation, I wanted to ask you, Nate, and then Alex: what quantum person, dead or alive, would you like to have dinner with?

Nate: Dead or alive. Well, that runs a risk. Can we just stick to dead, so that I don't have any apologies to make afterwards? I'm going to take that as a yes.

So I have to say, I grew up in physics reading Lev Landau, also known as the Course of Theoretical Physics. It's sitting across from me now. And somehow I just feel… I think Lev Landau actually died the year I was born, so we have a certain kind of soul transference or something.

So I think that's my answer.

Yuval: And Alex, how about you? And by the way, I know that you could always say dead and alive, but let's leave the superposition out of it for now.

Alex: I'm still thinking about who a quantum person is and what that looks like. I don't know, there are just so many. I guess the easy answer is to say Feynman, right? There are just so many insights that came from him, and from the information that you can find about him, he was such a great conversationalist, quantum or not. Just a fun person to be around.

But I don't know. Some of the people alive right now are fascinating to have conversations with about what the history of this field has been, right? And I think that's the thing that's, to me, so exciting about working in this field right now: things are moving so quickly that what is seen as the historical progression of quantum computing was developed by people who are still around and still active.

And of course there's Peter Shor, there's Umesh Vazirani. There are just so many people out there, and picking their brains is amazing. But I'm going to simplify it and say Feynman. He seems like a great guy to hang out with.

Yuval: So how can people get in touch with you to learn more about your work, and what type of people would you most like to hear from?

Alex: People can get in touch through our website. I hope that they're also going to the Braket website and starting to use our machines. Who do we want to hear from? Anyone who has great ideas, both about what they want to do with our devices and about where we should be moving the entire ecosystem. And people who are excited about joining forces and working together to make sure that this quantum computing adventure is successful.

Yuval: Perfect.

Well, thank you so much for joining me today.

Alex: Thank you. Have a great one.

Nate: Thank you so much.